Integration of multiple feature groups and multiple views into a 3D object recognition system

Abstract

This paper proposes two approaches for utilizing multiple-feature group and multiple-view information to reduce the number of hypotheses passed to the verification stage in an invariant feature indexing (IFI)-based object recognition system [8]. The first approach is based on a majority voting scheme that tallies the number of consistent votes cast by prototype hypotheses for particular object models. The second approach examines the consistency of estimated object pose from multiple scene-triples from one or more views. Monte Carlo experiments employing several hundred synthetic range images of objects in a large CAD-based 3D object database [7] show that a significant number of hypotheses can be eliminated by using these approaches. The proposed approaches have also been tested on real range images of several objects. A salient feature of our system and experiment design compared to most existing 3D object recognition systems is our use of a large object data base and a large number of test images.

This research was supported by NSF grants CDA-8806599 (to MSU), CDA-9121675 and IRI-9209212 (to WSU), and by the WSU Research Grant-In-Aid Program. ©1994 IEEE. Reprinted, with permission, from Proceedings of 2nd IEEE CAD-Based Vision Workshop, Champion, PA, February 1994, pp. 184–191.